Why screens at night feel different (and what your eye is doing)

It's a little before midnight. You pick up your phone in bed to check one thing. The screen is at the same brightness it was at three in the afternoon — same wallpaper, same icons. And yet it feels like someone turned the brightness up. The white app icons look piercing. The text under them looks slightly fuzzy, the way it doesn't during the day. You squint, scroll a little, and your eyes complain.

Nothing has changed about the screen. Something has changed about you. The room is dim, your visual system has spent the last hour or two adapting to that dimness, and now you're staring at a small, self-luminous object that is asking it to behave as though it's the middle of the day. The mismatch is the whole story. This post unpacks the parts of that story — pupil and adaptation, the (real but narrow) blue-light-and-sleep finding, dark mode, and what your contrast sensitivity is actually doing at night.

Your eye has adapted to the room, not the screen

In dim ambient light the visual system makes a set of slow adjustments. The pupil dilates from 2–3 mm to 4–6 mm. Neural gain is turned up — receptors and downstream ganglion cells become more sensitive to small luminance differences. After about ten minutes of low ambient light, the system is in a partly dark-adapted state: not full scotopic darkness, but not the photopic regime of a normally-lit day either. The mechanism details — pupil exposing lens aberration, photon-noise scaling, the rod-cone hand-off — are the same whether you're in a dim restaurant or scrolling in bed; they're covered in the low-light vision post.

What's specific to the screens-at-night case is the adaptation mismatch. A phone screen in a dark room is a small, bright, self-luminous object embedded in a much darker surround. Local retinal regions looking directly at the screen are being driven toward photopic adaptation — bright stimulus, high local mean luminance. The peripheral retina is still adapted to the dim room. The visual system has to mediate between two adaptation states at once, and it does so imperfectly. The screen looks brighter than the same luminance would at 3 p.m. because your sensitivity has gone up while the screen stayed put. White elements feel piercing because a 250 cd/m² icon hitting a partly-dark-adapted eye is, from the visual system's point of view, a much brighter stimulus than the same icon hitting a fully photopic one.

This is also why a phone in the dark produces a particular ache that the same phone outdoors does not. Being pinned at high gain while a bright object hammers a small patch of central retina is uncomfortable in a way the same screen at noon simply isn't.

What "dark mode" actually changes

Dark mode is the most popular consumer answer to the night-screen problem, and it is usefully a real reduction in emitted light. Inverting an interface replaces large areas of high-luminance white with low-luminance black or dark gray. On an OLED display the pixels in those regions emit essentially zero light. On an LCD the backlight is still there, but the visible result is still a much lower mean luminance.

That lower mean luminance is the headline benefit, and it's the part that matters for the adaptation-mismatch problem above. A dark-mode screen is more like the dim surround your eyes have adapted to. The peripheral and central retina are less at odds. The piercing-icon problem softens.

The rest of the dark-mode story is less clear-cut. Most of human reading history happens with dark text on a light background — books, newspapers, high-resolution print are all designed that way. Switching to light text on a dark background changes the ergonomic problem the visual system is solving. Studies on reading speed and comprehension have produced mixed results: some small advantages for dark-on-light at moderate luminance, no meaningful difference in others once readers acclimate, subjective preference effects that flip by age and time of day. Honest summary: for the screens-at-night case, dark mode helps because it lowers total emitted luminance and softens the surround mismatch. Whether it's a better reading surface in general is less settled.

A common pattern that does seem to work: dark mode for short bursts at night, light mode at lowered brightness for sustained reading. If you're doing forty minutes of reading at 11 p.m., the answer is usually "lower brightness and use a light theme," not "dark mode at maximum brightness."

The blue-light story is narrower than commonly told

Walk into any optical shop and someone will offer to sell you blue-light-blocking lenses with a story about retinal damage and screen-induced strain. The story is partly true and partly oversold. Two pieces have to be separated.

The mechanism is real. A specialised class of retinal ganglion cells — intrinsically photosensitive retinal ganglion cells (ipRGCs, cells at the back of the eye that respond to light directly rather than relaying signals from the rods and cones) — contain a pigment called melanopsin, most sensitive to short-wavelength light around 480 nm. These cells project to the suprachiasmatic nucleus, the brain's master circadian clock.¹ Bright short-wavelength light in the evening signals "daytime," suppresses melatonin, delays sleep onset, and shifts the circadian phase later. Well-established circadian biology.

The consumer-device claim is more mixed. Chang, Aeschbach, Duffy and Czeisler's 2015 study in PNAS compared four hours of pre-bedtime reading on a light-emitting iPad against the same time on paper under dim light, across five evenings. The iPad condition suppressed evening melatonin by about 55%, delayed dim-light melatonin onset (the evening rise in melatonin that marks the body clock's "night start") by more than 1.5 hours, reduced REM sleep, and left subjects sleepier the next morning.² The effect on circadian timing is real and measurable — but the iPad was held close to the face for four hours, a stronger exposure than a five-minute lock-screen check.

The often-repeated retinal-damage claim — that consumer screens emit enough blue light to harm the retina — is not supported at consumer-screen brightness. Phone and laptop screens emit orders of magnitude less short-wavelength power than midday outdoor sunlight; if our screens damaged retinas, a sunny afternoon would damage them more. The American Academy of Ophthalmology takes the same position: the small amount of blue light from digital devices has not been shown to damage the eyes.³

Lawrenson, Hull and Downie's 2017 systematic review in Ophthalmic and Physiological Optics reviewed the published evidence on blue-light-blocking spectacle lenses for visual performance, macular health, and sleep-wake cycle outcomes — and, with only three small trials meeting inclusion criteria, concluded there was a lack of high-quality evidence on all three.⁴ A larger 2023 Cochrane review of 17 trials reached a similar conclusion: blue-light-filtering lenses may not reduce eye strain with computer use (low-certainty evidence), probably make little or no difference to visual acuity, and left effects on sleep quality indeterminate, with no randomised evidence on macular protection.⁵ The mechanism (ipRGC → melatonin) is real; the consumer-product claim (these glasses fix your sleep / save your retina) is not well supported. Night-shift filters on your phone shift the screen's spectrum away from short wavelengths, which is mechanistically aligned with reducing ipRGC drive — but published evidence that this translates into meaningfully better sleep for the average user is thin.

The cleanest framing: if you want to defend your sleep, lower the brightness, put the screen away an hour before bed, and use a warm filter for the time you do use it. Brightness is the biggest lever; the spectral shift is plausibly a smaller, secondary help.

What contrast sensitivity does on a screen at night

Two distinct things happen to how well you see a screen as evening goes on.

Bright details against dark backgrounds become piercing. A white notification icon against a near-black phone background is a maximum-contrast stimulus. Your dark-adapted eye, sitting at high gain, registers that maximum contrast as harsher than the same stimulus would feel in daylight. The visual system has good contrast-gain control — it doesn't melt down — but the experience of high-contrast bright features on a dark interface is uncomfortable in a way it isn't at noon. This is one reason dark mode is sometimes less restful than expected: the dark surround is calmer, but the bright UI elements that remain are now sitting on a high-contrast pedestal.

Mid-contrast text gets harder to read. Body text on a screen — gray on white, off-white on dark — is mid-contrast. Reading it depends on the high-spatial-frequency arm of your contrast sensitivity function, the part of the CSF that does fine detail. Photopic contrast sensitivity is at its highest in well-lit daytime conditions; as ambient light drops and the visual system shifts toward partial dark adaptation, sensitivity to fine detail compresses. It is this fine-detail, high-spatial-frequency arm that is most fragile: Owsley, Sekuler and Siemsen's 1983 lifespan study found that sensitivity to coarse (low-spatial-frequency) patterns held roughly steady across adulthood, while sensitivity to fine detail fell off with age, and that much of the fine-detail deficit tracked how much light actually reached the retina.⁶ The net result on a phone at night: the same text that read crisply at 3 p.m. can feel slightly smudged at midnight, even if your acuity hasn't changed.

This isn't a problem with the screen. It's a feature of how your visual system spends its sensitivity at different times of day.

Practical adjustments

Reasonable interventions for a healthy adult who finds late-night screen use noticeably worse than daytime use. None of them are medical claims.

Lower the brightness in dim rooms. Most phones auto-adjust but often calibrate to a target that's too bright for a dark bedroom. Slide it down manually at night. A screen that just barely shows the content is comfortable; one that punches through the dark is not.

Use dark mode for short bursts, light mode for sustained reading. Lock-screen checks, messaging, brief scrolls: dark mode at lowered brightness. Forty minutes of an article: a light interface at lowered brightness is usually easier to read.

Make text bigger at night. This compensates directly for the reduced effective contrast sensitivity for fine detail. Even a one-step bump in system font size helps.

Have some ambient light. A dim lamp by the bed beats total darkness. It narrows the mismatch between screen and surround, keeps the eye in a less extreme adaptation state, and removes the piercing-icon problem almost entirely.

Turn on the night-shift filter — and don't expect it to fix your sleep on its own. The spectral shift is plausibly helpful for circadian timing; the dominant factor for sleep is still total light dose and timing.

If you wear glasses, ask about anti-reflective coatings. Internal reflections off the back surface of spectacle lenses, with a bright screen in front of the eye, contribute to a fraction of late-night screen discomfort.

What our test can — and cannot — tell you about night-screen fatigue

Note. A contrast sensitivity test is a screening signal of overall visual function, not a diagnosis of any specific condition. A lower-than-typical result is a reason to take a better question to your eye doctor — not a verdict. The night-screen experience this post describes is a normal physiological pattern; it does not require a diagnosis.

Our test runs in normally-lit indoor conditions and measures photopic contrast sensitivity — the daytime curve. It does not directly measure mesopic CSF, and it does not measure circadian or adaptation effects at all. What it does measure is the underlying photopic CSF your visual system carries into the night. People whose photopic CSF is below typical for their age band⁷ tend to find low-light and screen-at-night tasks harder than people whose photopic CSF sits closer to the population peak. The photopic measurement is an upper bound; everything else — dim rooms, dark adaptation, glare, fatigue — sits below that ceiling.

If your night-screen experience has been getting worse over months, taking the test once for a baseline and again in three to six months is a concrete way to check whether the underlying visual function has shifted versus the environment and habits. The self-tracking guide covers cadence and how to read a trend.

Take the test

Take the test now. Three minutes, a normally-lit room, a saved result. The calibration step is most of what separates "a measurement of you" from "a measurement of your laptop" — the screen-calibration post walks through the gamma and viewing-distance steps. For background on what contrast sensitivity measures, see the primer.